Gitarchana Roy

Levene’s test and Bartellet’s test is used to test if samples have equal variances. Certain tests like ANOVA , assume that variances are same across the sample. These tests can help in verifying the assumption. Similarities between Levene’s and Bartlett’s Test: 1. Both have the same Null and Alternate Hypothesis. i.e. Null = Variance are equal Alternate = Variances are not equal Differences Levene’s Test Bartlett’s Test Alternative to Bartlett’s test Alternative to Levene’s test Works if sample data is not normal. Assumes the sample data is normal This test is robust in case the data set has ordinal data. It uses the median of the sample data sets. It uses the mean of the sample data sets. Non parametric test Parametric test Example: Suppose a researcher is interested in comparing the effectiveness of three different trainings (A, B, and C) in reducing defect density. The defect density is measured as defects / hr, Higher defect density indicates less effective training. The researcher collects defect density data from three groups of participants: Group A (n = 30), Group B (n = 25), and Group C (n = 28). To assess the equality of variances, the researcher can perform both Levene's and Bartlett's tests based on data normality. The analysis can be as below: Ho = The variances between the 3 groups are equal Ha = The variances between the 3 groups are not equal If p value > 0.05, the variances between the 3 groups are equal and if p value < 0.05 , the variances between the 3 groups are not equal. If the data set is normal then the researcher should use Bartlett’s test and if the data set is not normal, the researcher should use Levene’s test.

IMR Control Charts is the most commonly used control charts which is a combination of the Individual and Moving range chart which helps us to analyse process stability and variability. It is used when we analyse continuous data and the data is normal. It is true that in the real world we seldom find normal data. In case of IMR charts, results will still be valid in case the data is moderately normal. The data can be transformed in case the data is non normal and highly skewed. The transformations can be done using either Box Cox Transformation, Logarithmic or Johnson’s Transformation. The data can be transformed and then appropriate control charts can be used to check on process stability Example:- 1. Usually time , effort and defect related data is non normal because the data can spread from 0 to infinite. One should investigate the data collection process and if MSA was done correctly before trying to transform data. Data transformation should be the last resort. Non parametric tests can also be used with non normal data. They usually analyse the median instead of the mean. However, it is possible that non parametric tests might not show the minor differences. In such cases data transformation is the only option left.

A Top-Down Diagram / Flowchart helps to visualize a process by listing down the main process steps and the sub steps listed below the main process steps. This helps to understand the entire process or problem at hand and identify the improvement opportunities. This can be used in the various DMAIC phases in a Six Sigma Project. Define phase to: · Define the problem in a hierarchical and structured manner · Understand the process / problem at hand Measure phase to: · Identify primary and secondary KPIs / measurements Analyze phase to: · Understand the areas of process improvement by analyzing all the processes and the sub processes. · Identify the most probable causes and conduct root cause analysis. Improve phase to: · Implement improvement actions in the necessary sub processes and processes so that the overall improvement can be analyzed. Control phase to: · Make necessary process changes and analyse it’s impact in a systematic manner.

A probability plot is a graphical technique for assessing whether the data set follows a normal distribution or not. The normal probability plot is also a type of Quantile plot. It is a graph used to assess the Process Capability i.e., if a straight line is plotted we say the process is normally distributed. The probability plot is always plotted by arranging data in the ascending order and creating quantiles. Insights from the graph Deviation from this straight line denotes deviation from normality. This usually happens if there are outliers in the data set. Skewness in the data set denotes asymmetrical data. Positive skewness denotes longer right tail and negative skewness denotes longer left tail. Kurtosis describes how much the probability distribution falls in the tails instead of its center. There can be 2 insights for kurtosis i.e., heavy tailed or Light tailed distribution. · If the curve bends upward more than expected in the upper quantiles or bend downward more than expected in the lower quantiles then it is a heavy tailed distribution or it has a higher probability of extreme values as compared to light tailed distributions. · If the curve bends downward more than expected in the upper quantiles or bends upward more than expected in the lower quantiles then it is a light tailed distribution i.e., lower probability of extreme values as compared to heavy tailed distributions.

Linear Regression Nonlinear Regression Represents relationship between variables with a straight line Represents relationship between variables with a curved line Example: Defects vs. Rework Example: Growth of Business i.e., Revenue with employee strength Form of linear model is typically either the constant or a parameter multiplied by an independent variable. Simple Addition. Rational function which is the ratio of 2 polynomial functions. R-squared value is valid R-squared value is invalid Might not capture true relationships if they are complex. Explains complex relationships Data set must be homogeneous. Might be overlooked while creating models. Better fit and prediction accuracy Easy to understand. Difficult to interpret and comprehend results. Governing Criteria: If better model fit is essential, then nonlinear regression should be selected. If simple, easy to understand models need to be created then Linear models should be created. If prediction accuracy is important, then Nonlinear regression should be selected.

A Job Breakdown Sheet is a tool to breakdown the job into multiple tasks, sub tasks and activities in a hierarchical / tree like structure. It is usually undertaken during the planning phase of the project so that each activity can be thought through and listed down. Effort and No. of people required to complete each task along with the timelines is estimated. Finally, it helps us to derive the overall cost of the project. Advantages Disadvantages Enables streamlining tasks and understanding dependencies. Time Consuming – It takes time to create a JBS although tools are available to expedite it. Enables monitoring and Control – It can be placed in a common server or a common place accessible to team members and supervisors. Status can be tracked against each task and proactive actions can be taken for due course correction. This tool can be used during the Measure and Control phase of Lean Six Sigma DMAIC framework as it helps in monitoring the progress and taking timely course corrections. It is expected to be updated regularly and kept live so that it can throw real time issues. It can be used to provide weekly updates to Clients.

OCAP is a structured approach to ensure data correction so that the process under study operates within the Specification limits i.e., Upper and Lower Specification limits which is the acceptable operating range. It is similar to the concept of Outlier analysis used in Control Charts wherein we analyze the root cause of the data points that exceed the specification limits. It involves the below steps: 1. Define the problem statement for the given process. 2. Identify Root cause using either 5 Why analysis or Fishbone analysis. Pareto charts can be used in case we are analyzing defects. 3. Identify the corrective actions and prioritize them in case of multiple actions. 4. Identify the preventive actions so that similar problems do not occur in future. 5. Implement the CAPA (Corrective and preventive actions). Identify owners and monitor the action to closure. 6. Continue monitoring the process and monitor if the implemented actions were effective. This can be checked if there is any process improvement. 7. This ultimately leads to process stability and a stable process is always predictable. 8. This also leads to customer satisfaction through better services.

Workload Balancing is a concept which is predominantly applied in any industry to ensure equal distribution of work across team members to increase productivity ,reduce risk of not meeting deadlines and reduce idle time. Manufacturing Industry Service Industry Capacity E.g., In car manufacturing industry with different departments undertaking objective of manufacturing different parts of the car. Workload balancing depends on the capacity of the equipment to deliver each unit of product. Capability E.g., In AI company workload balancing depends on the skill sets available in the organization to deliver as per client expectations. 1. RPA can help in automating repetitive tasks like scheduled downloads, refreshes etc. This saves time and effort and team can focus on value added activities. 2. AI and Machine Learning can help in analyzing past data and forecasting Utilization and FTE i.e., Increase Utilization and predict FTE availability by monitoring workloads. It helps to monitor allocation i.e., people who are underutilized and assigning them work with Clients where there is demand. Overall workload balancing inculcates unbiasedness through equal distribution of work and boosts employee morale.

Overproduction and Overprocessing are 2 categories of Lean wastes that Organizations most commonly use in their journey towards lean. Service: 1. Overprocessing Exhaustive testing practices might not be required for all types of project. Ideally , we should do a thorough testing prior to any deliverable to the client. Typically, in Data Engineering projects thorough Unit Testing, System Testing, Integration and Acceptance Testing including regression and smoke testing is recommended. There are consulting and predictive Modelling projects where exhaustive testing is not required. It is ok to have Unit testing done with Client provided test data followed with interim client demos and capture client review comments. Exhaustive testing will lead to Overprocessing. 2. Overproduction Release of processes that are not required by the project teams leads to overproduction and loss of time and effort in creating, reviewing and releasing them. For e.g. a start-up company which has embarked on the journey of process standardization might not need a tailoring and deviation process. This waste can be handled by waiting for the organization to mature and standardize so that standardization is a way of life and deviating from it will need process to be followed. Manufacturing: 1. Overproduction In Bakery shops, baking and preparing cakes without demand leads to overproduction. Cakes which have icing on it are such items that do not last beyond a day or 1.5 days. Hence, overproduction leads to waste of money, effort and time. The seller should forecast based on the time of the year, i.e. probable marriage dates etc. and then over produce. This can be handled by having tie ups with bigger hotels wherein beyond a certain time i.e. 10pm the cakes etc. are delivered to them. Bigger brands typically have a higher rate of footfall than the smaller joints. 2. Overprocessing Tailor creating clothes - The tailor should add his/her creativity to the extent that is required by the client else the clothes will appear overdesigned and will not fit the purpose. Such waste can be handled by waiting for the right occasion where the clothes are befiting.